Post on 14-Apr-2022
Internat. J. Math. & Math. Sci.
VOL. 14 NO. (1991) 191-202191
ON THE WEAK LAW OF LARGE NUMBERS FOR NORMED WEIGHTED SUMSOF I.I.D. RANDOM VARIABLES
ANORI AOLER
Department of MathematicsIllinois Institute of TechnologyChicago, Illinois 60616 U.S.A.
ANDREW ROSALSKY
Department of StatisticsUniversity of Florida
Gainesville, Florida 32611 U.S.A.
(Received March 14, 1990)
nABSTRACT. For weighted sums a:Y: of independent ancJ identically.distributed random variables
,IJ (n l/=1 p{Yn, n>l}, a general weak law of rrge numbers of the form :- na Y v bn ---,0 ts estabhshed where
1J\= /{Vn, n_> 1} and {bn, n>_ 1} are statable constants. The hypotheses revolve botli the behavior of the tad of the
distribution of IYll and the grOWnbehavlors of the constants {an, n_> 1} and {bn, n_> 1}. Moreover, a weak
law is proved for weighted sums a;Y; indexed by random variables {Tn, n>_ 1}. An example is presented"1
wherein the weak law holds but tl’trong law fails thereby generalizing a classical example.
KEY WORDS AND PHRASES. Weighted sums of independent and identically distributed random variables,
weak law of large numbers, convergence in probability, random indices, strong law of large numbers, almost
certain convergence.
1980 AMS SUBJECT CLASSIFICATION CODES. Primary 60F05; Secondary 60F15.
I. INTRODUCTION.
Let (Y, Yn, n_ I} be independent and identically distributed (i.i.d.) random variables defined on a
probability space (f/,s$,p), and let (an, n_1}, (Vn, n_1}, and {bn, n_l} be constants with an0, bn>0,
n_l. Then (anYn, n_l} is said to obey the general weak law of large numbers (WLLN) with centering
constants (vn, n_l} and norming constants (bn, n_1} if the normed and centered weighted sum
a.Y. n)/bn has the weak limiting behaviorj=l
n
j=t 0 (1.1)bn
where denotes convergence in probability. Herein, the main result, Theorem 1, furnishes conditions on
{an, n_>l}, {bn, n_>l}, and the distribution of Y which ensure that {anYn, n>_ 1} obeys the WLLN (1.1) for
suitable {vn, n_>l}. It is not assumed that Y is integrable. Of course, the well-known degenerate
convergence criterion (see, e.g., Loire [1, p. 329]) solves, in theory, the WLLN problem. The advantage of
employing Theorem lies in the fact that, in practice, its conditions (2.1), (2.2), and (2.3) are simpler and
more easily verifiable than the hypotheses of the degenerate convergence criterion. Jamison et al. [2] hadn
investigated the WLLN problem in the special case where an>0, bn a;, n> 1, and max a. O(bn).j=l _<j_<n
192 A. ADLER AND A. ROSALSKY
Conditions for {anYn, n>_ 1} to obey the general strong law of large numbers (SLLN)
n
j--1bn 0 almost certainly (a.c.)
had been obtained by Adler and Rosalsky [3,4]. In Section 5, an example illustrating Theorem is presented
and the corresponding SLLN is shown to fail.
The WLLN problem is studied in Theorem 2 in the more general context of random indices. More
specifically, let {Tn, n_> 1} be positive integer-valued random variables and let 1_<an--,o be constants such
that P{Tn/an>A} o(1) for some A>0. Theorem 2 provides conditions for
Tnj=l V[Cn] 0,
b[an]where the symbol Ix] denotes the greatest integer in x.
As will become apparent, Theorem 2 of Klass and Teicher [5] and Theorem 5.2.6 of Chow and
Teicher [6, p. 131] provided, respectively, the motivation for Theorems and 2 herein. Moreover, our
Theorems and 2 are proved using an approach similar to that of the earlier counterparts.
Some remarks about notation are in order. Throughout, a sequence {Cn, n>_l] is defined by
cn bn/lanl, n_> 1, and the symbol C denotes a generic constant (0<C<oo) which is not necessarily the
same one in each appearance. The symbols Unl" or Unl are used to indicate that the given numerical sequence
{Un, n>_ 1} is monotone increasing or monotone decreasing, respectively.
2. A PRELIMINARY LEMMA.
The key iemma in establishing Theorems and 2 will now be stated and proved. It should be noted
that the conditions (2.2) and (2.3) are automatically satisfied for the standard assignment of an-l, bn--nn>l.
LEMMA. If
nP{lYl>cn} o(I) (2.1)
and either
then
al2EY2I([YI _<Cn) o(bn2).j=l
PROOF. Note at the outset that cn’[" under either (2.2) or (2.3) and that (2.3) ensures
2 19.\a. obfi). (2.4)
j=l
Thus, (2.4) holds under either (2.2) or (2.3).
{Bnk, 0_<k_<n, n_> 1} by
Let c0=0 and dn=cn/n n>_l. Define an array
WEAK hAW OF LARGE NUMBERS FOR NORMED WEIGHTED SUMS 193
2 2
nnj=l J\ k
0 for k=0, n,n>l.
for l<k<n-1, n>2
and
It will now be shown that {Bnk 0_<k_<n, n_>l} is a Toeplitz array, that is,
n
k:__oIBnkl 0()
Bnk0 as n--,o for all fixed k>0.
Clearly (2.4) entails (2.6). To verify (2.5), note that Bnk>0, 0<k<n, n>l, since Cn’. Now under (2.2), for
all n>2,
n 2’/n-1 ((k+l)2-k2)d(< (12 j__laj )lkk=bn k ) (since dn |)
2/n_1 2\<- 3n a ldk =O(1)
and so (2.5) holds. On the other hand, under (2.3),
n 2dn" and a. < Cna2n n>l.j=l
Then for all n > 1,
Thus for all n >_ 2,
n
j=l Cn C_C__
n (_(nl (k+l)2d2k+l-k2d)Bnk <k=0 \ndnJ\k=l k
\nd/\k=l"
C2 ((k+l)dk+l + C ),k__ldk+,)Cnd2n 2C(n-1)d2n
(since tint)-< + nd2n0(1)
and again (2.5) holds thereby proving that {Bnk, 0<k_<n, n>_ 1} is a Toeplitz array.
194 A. ADLER AND A. ROSALSKY
Then by (2.1) and the Toeplitz lemma (see, e.g., Knopp [7, p. 74] or Love [1, p. 250]),
Next, note that
n
k=0BnkkP{IYl>Ck o(1). (.7)
I ai2Ey21([Y[<cn)
a2 Ey21(Ck <lVl<Ck)2n2 j=l k=l
j--’’l k’--1 x x- IYI <ck
_i n n
n n-I 21_.. _.laj2(cp{.y,>0}. c2nP{,y.>cn} + k__l(Ck+l_C)p{,y.>ck})b2n j_-
(by the Abel summation by parts" hmma)
n 2 n1/c.4-1- k<bn j=l k’--l k
nE BnkkP{lY[>ck} + o(1)
o(I) (by (2.7))
thereby proving the Lemma. []
3. THE MAIN RESULT.
With the preliminaries accounted for, Theorem may be stated and proved. As was noted in the
proof of the Lemma, the hypotheses to Theorem entail (2.4) and so necessarily bn--oo. However, it is not
assumed that {bn, n_> 1} is monotone. (In most SLLN results, monotonicity of {bn, n> 1} is assumed.)
THEOREM 1. Let {Y, Yn, n>l} be i.i.d, random variables and let {an, n>l} and {bn, n>l} be
constants satisfying an0, bn>0 n>l, and either (2.2) or (2.3). If (2.1) holds, then the WLLN
n (yj )Eaj=l EYI(IYI-<Cn)(3.1)-.P0
bn
obtains.
PROOF. Define Ynj Y.iI(IYj I-<cn), l_j_n, n_ I. For arbitrary >0,. ai(Y;-Ynj)P/]J=1" -n >} <PJ[-Y.:f:Ynj]}<nP{[Y,>cn}=O(1)(by(2.1)).[j=t_.,
WEAK LAW OF LARGE NUMBERS FOR NORMED WEIGHTED SUMS 195whence
Also,
since for arbitrary >0,
J= -.P0.bn
j=lbn
P
(3.2)
(3.3)
n1.._[_ [ ai2Ey21(iYl<cn) o(1)> _< e2b2nj=lJ
by the Lemma. The conclusion (3.1) follows directly from (3.2) and (3.3). E!
REMARKS.n 2" a. O(na2n), then
j=l
(i) Apropos of the condition (2.2), if cn/n is slowly varying at infinity and
orn 2
j=lj=l
PROOF. Note thatn 2a.
j=l Cna2n C(n/cn)2
o(I)b2nby slow variation (see, e.g., Seneta [8, p. 18]). Then slow variation yields (see, e.g., Feller [9, p. 281])
o( 3 o
j=l
(ii) Adler [10] provM a rtial conver of Theorem 1.
(iii) In the spirit of Kl and Teicher [5], Adler [11] h employ Threm to obtain a generaliz
one-sid law of the iterat logarithm (LIL) for weight sums of i.i.d, random variabl barely with or
without finite mean thereby generMizing me of the work of [5]. (Corollary low h n obtain by
KI and Teicher [5] and they u it in their invtigation of the LIL for i.i.d. ymmetric random
variabl.) To mewhat more scific, Adler [11] employed the WLLN (3.1) to obtain the a.c. limitingn
value of me (nonrandom) subquence of Yj/bn thereby yielding an upr und for the a.c. value ofn
lim inf a.Y./bn.nj=lThe ensuing Corollary is a WLLN analogue of Feller’s [12] famous generalization of the
Marcinkiewicz-Zygmund SLLN.
COROLLARY (KI and Teicher [5]). Let {Y, Yn, nl} i.i.d, random variabl and let
{bn, nl} itive constants such that either bn/n or
2
bn’, 1, bn 1(), and O (3.4)
Then nE Y" nEYl(IYl<bn)j=l
bn0 iff nP{lYl>bn} o(I).
196 A. ADLER AND A. ROSALSKY
PROOF. Sufficiency follows directly from Theorem whereas necessity follows from the degenerate
criterion noting that the family {(Yj- EYl(IYl__bn))/bn, l_<j_n, n_l} is uniformlyconvergence
asymptotically negligible. []
REMARK. In the Klass-Teicher [5] version of Corollary 1, the second condition of the assumption
(3.4) appears in the stronger form bn/n0.
The next corollary is an immediate consequence of Corollary and is the classical WLLN attributed
to Feller by Chow and Teicher [6, p. 128].
COROLLARY 2. If {Y, Yn, n_> 1} are i.i.d, random variables, then
n
j=IYj VnP-0n
for some choice of centering constants {vn, n_> 1] iff
nP{IYl>n} o(1).
In such a case, vn/n EYI(IYI_<n + o(1).
The next corollary removes the indicator function from the expression in (3. I).
COROLLARY 3. Let {Y, Yn, n_>l} be i.i.d. L random variables and let {an, n>_l} and {bn, n>_l}n
be constants satisfying an:/:0 bn>0 n>l, and either (2.2) or (2.3). If (2.1) holds and M lim a./bnn--,oj=
exists and is finite, then
n
J.’.-,l _.P M(EY).bn
PROOF. First observe that (3.1)obtains by Theorem 1. Now n--,oolimCn oo since a2n o(b2n) by
(2.4). Then by the Lebesgue dominated convergence theorem, EYI(IYI_<Cn)--, EY, whence
nE a;EYl(lYl_<Cn)j=l
bn M(EY)
which when combined with (3.1) yields the conclusion. []
4. A WLLN WITH RANDOM INDICES.
In this section, Theorem is extended to the case of random indices {Tn, n_> 1}. No assumptions are
made regarding the joint distributions of {Tn, n_ 1} whose marginal distributions are constrained solely by
(4.1). Moreover, it is not assumed that the sequences {Tn, n_l} and {Yn, n_l} are independent of each
other. It should be noted that the condition (4.1) is considerably weaker than Tn/cn-.Pc for some constant
c[0,oo).
THEOREM 2. Let {Y, Yn, n_l}, {an, n_l}, and {bn, n_l} satisfy the hypotheses of Theorem
and let {Tn, n_l} be positive integer-valued random variables and l_an--.oo be constants such that for
some A>O
and
b[,n] O(b[,n]) if >1 (4.2)
WEAK LAW OF LARGE NUMBERS FOR NORMED WEIGHTED SUMS 197
hold. Then
TIIj-laj(Yj-= Ynj) _p O.
b[an]
For arbitrary >0 and all large n,
t b[an]> cJ
Tn Tn< P X a.Y. # __lajYnj
<_ P U I11>[. +o()t j=l t-
(by (4.1))
=(! + o(,))A[Cln]P{iYi><[ttn]} +o(1)
=o(1) (by (2.1))
thereby establishing (4.3).
Thus, to complete the proof, it only needs to be demonstrated that
Tna.(Y. EYnj)J=lJ "J P0.
(4.3)
(4.4)
To this end, for arbitrary >0 and all large n,
EYnj)Ip]j=lt %5 > }
198 A. ADLER AND A. ROSALSKY
(by (4.1))
=P max a. Y .-EYnj > cb[an] +o(1)_<k_<[=n] = n
-< )+[n] = (by the Kolmogorov inequality)
if O<A<I
if A> (by (4.2) and Cn ]’)
o(1) (by the Lemma)
thereby establishing (4.4) and Theorem 2. []
REMARKS. (i) The referee to this paper so kindly supplied the following example which shows that
Theorem 2 can fail if the norming sequence {b[an], n_l) is replaced by {Tn, n_l}. Let {Y, YH, n>l} be
i.i.d. Cauchy random variables and let
an=l bn=nI+, Tn=n, an=n, n>l
where >0. Then (2.1) and (2.3) hold and trivially Tn/an--Pl and hence the conclusion to Theorem 2
obtains, but
[ aj Yj EYI(IYI <Clan] E Y" pj=l J- -/. O.Tn
(ii) The referee also suggested that the authors look into the question as to whether in Theorem 2 the
norming sequence can be taken to be _{bTn’ n_l_}. The ensuing corollary provides conditions for the answer
to be affirmative. It should be noted that the pair of conditions (4.1) and (4.5) is equivalent to the single
condition
which is clearly weaker than Tn/an P c for some constant 0<c<oo.
COROLLARY 4. Let {Y, Yn, n_>l}, {an, n_>l}, {bn, n_>l}, and {Cn, n_>l} satisfy the hypotheses
of Theorem 2 and suppose, additionally, that bnT and for some At>0 that
WEAK LAW OF LARGE NUMBERS FOR NORMED WEIGHTED SUMS 199
and
pTn ,,} o(1) (4.5)an <
b[anl O(b[Atanl) if A’<I (4.6)
hold. Then
j=l
bTnPROOF. In view of Theorem 2, it suffices to show that b[anl/bTn., is bounded in probability, that is,
for all >0, there exists a constant C<oo and an integer N such that for all n>_N
(4.7)
To this end, let >0. If At_>l, then letting C=I, the monotonicity of {bn, n_>l} guarantees that
b[an] < Cb[A,an], n>_ (4.8)
whereas if At<l, then (4.6) ensures (4.8) for some constant C<oo. Thus, (4.8) holds in either case. Then for
all large n,
<_ P{[b[an]>CbTn’] [Tn >[A’an]]} +
_< P{b[an]>Cb[Atan]} + (by bn]’ and (4.5))
thereby establishing (4.7) and Corollary 4.
(iii) The ensuing example shows that, in general, Theorem 2 can fail if the norming sequence
{b[an], n_>l} is replaced by {bTn n>l}. Let {Y, Yn, n>_l} be i.i.d, random variables with Y having
probability density function
l[e,oo)(y), -oo<y<oof(Y) ;21oCg ywhere C is a constant and let
an=l bn=n Tn:[q-6, an=n, n_> 1.
Now for all n_>3, employing Theorem of Feller [9, p. 281],
oo
nP{[Y[>n} nCn y210g y
dy(I+o(1))C
log n o(1).
All of the hypotheses to Theorem 2 are satisfied and hence the conclusion to Theorem 2 obtains. Assume,
however, that
200 A. ADLER AND A. ROSALSKY
Tnj=l
aTn_J=
vj-P0 (4.9)
prevails. Then
n
j=ln EYI([Y[ <n2) P
But by Corollary 2,
n
j=l EYl(lYl<n) 0.n
whence via subtraction EYI(n<IY[_<n2) o(1). But for n_>3,
EYl(n<[Y[<n2) /n2 C dy C(,og log n2- log log n) C log 2,n y log’y
a contradiction. Thus, (4.9) must fail.
The last corollary of this section, Corollary 5, is a random indice version of the sufficiency half of
Corollary 2, and it is Theorem 5.2.6 of Chow and Teicher [6, p. 131]. Corollary 5 follows immediately from
Corollary 4 by taking an-1 bn-n, an--n, n>_l.
COROLLARY 5. Let {Y, Yn, n_>l} be i.i.d, random variables such that nP{[Yl>n} o(1) and let
{Tn, n_> 1} be positive integer-valued random variables such that
Tn -, c for some constant 0<c<oo.
Then
TRj=l J_ EYI([YI<n) 0.Tn
5. AN INTERESTING EXAMPLE.
In this last section, a generalization of a claical example is presented. A sequence of weighted i.i.d.
random variables {anYn, n>l} is shown, via Theorem 1, to obey a WLLN. On the other hand, the
corresponding SLLN is shown to fail. It should be noted that ElY cw. The classical example is the special
case 6= and an 1.
EXAMPLE. Let {Y, Yn, n_>l} be i.i.d, random variables with Y having probability density function
C61yl)6 I(_f(y)y2(los oo,_e]U[e,oo)(y),
where 0<6_(1 and C6 is a constant. Then for every sequence of constants {an, n_) 1} with
nE a.Y.
(5.1)J= P 0,
but
WEAK LAW OF LARGE NUMBERS FOR NORMED WEIGHTED SUMS 201
n n
j-1 j=llira sup -lira inf
and, consequently, for any constant cE(-oo,cx)n
j=lPn li-*moo nlan[ =c =0.
PROOF. Set bn’-nlanl n>l. Then cn=n n>l, and both (2.2) and (2.3) hold. Now for all n_>3,
employing Theorem of Feller [9, p. 281],
nP{[Y[>n} 2nC6 dy o(1),n y2(log y)b (log n)
and so (5.1) follows from Theorem since EYI(IYI<_n) 0, n>_ t.
Next, for arbitrary 0<M<oo, E oo ensures that
ooP Iynl
,whence by the Borel-Cantelli lemma
Plim sup ’Yn__J > M}> P{l > M i.o.(n)/ 1.
Since M is arbitrary,
n.Y.n-1
I]=1 j=loo lim sup :’-’ lim sup
n --oc n --oo nlanl
and so
< lira sup + lim sup(n-l) a.c.,
no nlanl n--o lan_l
implying (5.2) via symmetry and the Kolmogorov 0-1 law. E!
ACKNOWLEDGEMENT. The authors wish to thank the referee for supplying the first example and for
posing the question concerning normalization by _{bTn n_>l}_ which were presented in the Remarks following
the Proof of Theorem 2. The authors are also grateful to the referee for her/his very careful reading of the
manuscript and for some comments which enabled them to improve the presentation of the paper.
REFERENCES
1. LOIVE, M. Probability Theory, Vol. I, 4th ed., Springer-Verlag, New York, 1977.
JAMISON, B.; OREY, S. and PRUITT, W. Convergence of Weighted Averages of Independent RandomVariables, Z. Wahrsch. verw. Gebiete 4 (1965), 40-44.
202 A. ADLER AND A. ROSALSKY
3. ADLER, A. and ROSALSKY, A. On the Strong Law of Large Numbers for Normed Weighted Sums ofI.I.D. Random Variables, Stochastic Anal. Appl. fi (1987), 467-483.
4. ADLER, A. and ROSALSKY, A. On the Chow-Robbins "Fair" Games Problem, Bull. Inst. Math.Academia Sinica 17 (1989), 211-227.
5. KLASS, M. and TEICHER, II. Iterated Logarithm Laws for Asymmetric Random Variables BarelyWith or Without Finite Mean, Ann. Probab. (1977), 861-874.
6. CHOW, Y.S. and TEICHER, H. Probability Theory: Independence, Interchangeability, Martingales,2nd ed., Springer-Verlag, New York, 1988.
7. KNOPP, K. Theory and Application of Infinite Series, 2nd English ed., Blackie and Son, London, 1951.
8. SENETA, E. Regularly Varying Functions, Lecture Notes in Math. 508, Springer-Verlag, Berlin, 1976.
9. FELLER, W. An Introduction to Probability Theory and Its Applications, Vol. II, 2nd ed., John Wiley,New York, 1971.
10. ADLER, A.B. Some Limit Theorems for Weighted Sums of Random Variables, Ph.D. dissertation,Dept. of Statistics, Univ. of Florida, Gainesville, Florida, 1987.
11. ADLER, A. Generalized One-Sided Laws of the Iterated Logarithm for Random Variables BarelyWith or Without Finite Mean, J. Theoretical Probab. . (1990).
12. FELLER, W. A Limit Theorem for Random Variables with Infinite Moments, Amer. J. Math. 68(1946), 257-262.
Submit your manuscripts athttp://www.hindawi.com
Hindawi Publishing Corporationhttp://www.hindawi.com Volume 2014
MathematicsJournal of
Hindawi Publishing Corporationhttp://www.hindawi.com Volume 2014
Mathematical Problems in Engineering
Hindawi Publishing Corporationhttp://www.hindawi.com
Differential EquationsInternational Journal of
Volume 2014
Applied MathematicsJournal of
Hindawi Publishing Corporationhttp://www.hindawi.com Volume 2014
Probability and StatisticsHindawi Publishing Corporationhttp://www.hindawi.com Volume 2014
Journal of
Hindawi Publishing Corporationhttp://www.hindawi.com Volume 2014
Mathematical PhysicsAdvances in
Complex AnalysisJournal of
Hindawi Publishing Corporationhttp://www.hindawi.com Volume 2014
OptimizationJournal of
Hindawi Publishing Corporationhttp://www.hindawi.com Volume 2014
CombinatoricsHindawi Publishing Corporationhttp://www.hindawi.com Volume 2014
International Journal of
Hindawi Publishing Corporationhttp://www.hindawi.com Volume 2014
Operations ResearchAdvances in
Journal of
Hindawi Publishing Corporationhttp://www.hindawi.com Volume 2014
Function Spaces
Abstract and Applied AnalysisHindawi Publishing Corporationhttp://www.hindawi.com Volume 2014
International Journal of Mathematics and Mathematical Sciences
Hindawi Publishing Corporationhttp://www.hindawi.com Volume 2014
The Scientific World JournalHindawi Publishing Corporation http://www.hindawi.com Volume 2014
Hindawi Publishing Corporationhttp://www.hindawi.com Volume 2014
Algebra
Discrete Dynamics in Nature and Society
Hindawi Publishing Corporationhttp://www.hindawi.com Volume 2014
Hindawi Publishing Corporationhttp://www.hindawi.com Volume 2014
Decision SciencesAdvances in
Discrete MathematicsJournal of
Hindawi Publishing Corporationhttp://www.hindawi.com
Volume 2014 Hindawi Publishing Corporationhttp://www.hindawi.com Volume 2014
Stochastic AnalysisInternational Journal of